FDE in Japan

Event Introduction

For more details, please visit: https://www.tokyo-ai.org/events/evt_1024_2026
Generative AI is entering the enterprise, but moving AI from demos into real business operations requires more than models and products. It also requires Forward Deployed Engineers (FDEs): engineers who understand both technology and business, work closely with customers, and turn complex needs into practical solutions. From Palantir to OpenAI, FDEs are becoming an increasingly important role in the AI era.

In Japan, however, FDE is still a relatively new career path. What does an FDE actually do? How can a software engineer transition into FDE? What capabilities does an excellent FDE need? How do FDEs work with Japanese companies?

At this event, four speakers from different backgrounds will discuss FDE from four perspectives: career transition, technical practice, customer-facing delivery, and enterprise AI implementation in Japan. If you work in AI, software development, or consulting, or are considering your next career direction, this is an opportunity for practical exchange and discussion.

Suitable for:

• Software engineers
• AI / ML engineers
• Solutions engineers
• Solutions architects
• Technical consultants
• Product managers
• IT consultants
• Engineers considering a transition into FDE
• Entrepreneurs exploring enterprise AI implementation
• DX / AI leaders at Japanese companies

Event Agenda
13:30 – 14:00 | Check-in & open networking
14:00 – 14:10 | Opening: Why Japan needs FDEs now
14:10 – 14:40 | How to Become an FDE
In the age of AI, what is truly scarce is no longer simply the ability to write code, but the ability to work closely with customers, identify real problems, and bring AI into real business operations. This session explores how to become an FDE, clarifying how the role differs from that of a software engineer, implementation specialist, solutions architect, or consultant. It also explains how to move beyond taking requirements and delivering projects toward discovering problems, creating business outcomes, and building reusable product capabilities. Whether you come from research and development, implementation, pre-sales, product, or consulting, you can find a suitable path to transition into FDE.
Speaker: Suwei Gui — Director, Tokyo Generative AI
Microsoft MVP (AI), founder of the Tokyo Generative AI Development Community, author of the technical blog Gui Ji, and CTO of elifes system Inc. He leads the development of the AI business social platform Linkora.
14:40 – 14:50 | Q&A & discussion
14:50 – 15:20 | What AI capabilities should an FDE have?
In the AI era, what an FDE needs is not just development capability, but the ability to understand the boundaries of AI capabilities and translate technology into real business value.
This session explores what kinds of AI capabilities an FDE should have, drawing on practices related to enterprise AI adoption, AI Agents, and Robotics. It examines how FDEs can understand business problems, judge whether a technology is suitable, and make appropriate technical choices amid complex systems and real-world constraints.
As AI evolves from single models into intelligent systems that connect data, systems, and the physical world, the FDE role continues to change. More important than mastering a specific technology is building a broad understanding of AI, systems thinking, and sound judgment, then bringing technology into real-world scenarios.
Speaker: Louis Lee — Arvione AI Inc. | CEO / AI Transformation Architect
Leads enterprise AI Transformation from concept to implementation and adoption, with Generative AI, AI Agents, and Robotics at the core. Provides end-to-end support across business problem structuring, AI technology strategy, system architecture and integration, PoC evaluation, production deployment, and continuous improvement. His strength lies in connecting advanced technology to tangible business outcomes.
15:20 – 15:30 | Q&A & discussion
15:30 – 16:00 | AI Does Not Fix Chaos: What an FDE Should Stop Doing
“Please use AI to somehow solve the most troublesome business process”—projects usually begin with this request. And the most troublesome business is often the most chaotic: manuals do not match actual operations, exceptions make up half of the work, and the real rules exist only in one person’s head. Putting AI into such a process does not cure the chaos. It only automates it—and makes it faster and bigger.
I see an FDE’s work as having four parts: untangling chaotic business processes, distinguishing where AI is and is not suitable, implementing the result as a system in the field, and keeping it running. The fourth part is why the second part is necessary—when too much AI is added, the system will not keep running.
In this session, I will use a real project to recreate the second part in full: the initial plan I drew, the moment my thinking changed in the field, the places where we ultimately decided not to use AI, and the results. I will also discuss the parts where my judgment was wrong. This is not a career-path talk; I will break down the actual work itself.
Speaker: Jun Li — Principal Forward Deployed Engineer, UiPath Inc.
UiPath’s first FDE in Japan, with experience supporting more than 100 customer sites as an advisor on automation and AI. He specializes in untangling complex business processes and turning them into systems that continue to operate in the field.
16:00 – 16:10 | Q&A & discussion
16:10 – 16:40 | Practical AI Agent Development with Salesforce Headless360 and Context Build
Speaker: He Jian — Salesforce Japan Lead Forward Deployed Engineer
Shares a practical case study on enterprise AI Agent development using Salesforce Headless360 (sf-pi) as the foundation for AI Agent skills and advancing application development through a Context Build approach.
Instead of asking AI to write code directly from raw requirements, the team first used AI to organize requirements into structured specifications, design drafts, open questions, and implementation plans. Human review then determined what could enter the formal context. sf-pi supported Salesforce querying, validation, deployment, and evidence recording.
The team also improved issues such as instruction bloat, oversized context, and repeated reviews through checkpoints, role separation, and context compression. The focus is not simply making AI write code, but establishing a sustainable, traceable, and verifiable AI Agent development process.
16:40 – 16:50 | Q&A & discussion
16:50 – 17:40 | Networking
17:40 | Event ends